Multi-regional Clinical Trials with Heterogeneous Regional Treatment Effects

In this article, the authors first review the fixed-effects model (FEM) and illustrate the assumption violation of FEM in multi-regional clinical trials (MRCTs) when regional differences are anticipated. The authors illustrate the definition of the overall treatment effect and further investigate the insights of discrete random effects model (DREM) to understand the estimation of the overall treatment effect in an MRCT. The authors focus on two practical issues: (1) the evaluation of an MRCT with unbalanced patient allocation and (2) the sample size re-estimation based on interim data of an MRCT. For the first issue, the authors interpret the relationship between Fisher information and effective sample size, so that the design and data analysis of an MRCT with an unbalanced patient allocation can be evaluated in a similar manner to balanced patient allocation through effective sample size. For the second issue, the authors illustrate the possible bias of parameter estimates and Z-score statistics due to sample size re-estimation at an interim time of an MRCT. The authors further proffer a solution to adjust the biased test statistic for testing the overall treatment effect.